1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
|
SUBROUTINE SLAQPS( M, N, OFFSET, NB, KB, A, LDA, JPVT, TAU, VN1,
$ VN2, AUXV, F, LDF )
*
* -- LAPACK auxiliary routine (version 3.3.1) --
* -- LAPACK is a software package provided by Univ. of Tennessee, --
* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
* -- April 2011 --
*
* .. Scalar Arguments ..
INTEGER KB, LDA, LDF, M, N, NB, OFFSET
* ..
* .. Array Arguments ..
INTEGER JPVT( * )
REAL A( LDA, * ), AUXV( * ), F( LDF, * ), TAU( * ),
$ VN1( * ), VN2( * )
* ..
*
* Purpose
* =======
*
* SLAQPS computes a step of QR factorization with column pivoting
* of a real M-by-N matrix A by using Blas-3. It tries to factorize
* NB columns from A starting from the row OFFSET+1, and updates all
* of the matrix with Blas-3 xGEMM.
*
* In some cases, due to catastrophic cancellations, it cannot
* factorize NB columns. Hence, the actual number of factorized
* columns is returned in KB.
*
* Block A(1:OFFSET,1:N) is accordingly pivoted, but not factorized.
*
* Arguments
* =========
*
* M (input) INTEGER
* The number of rows of the matrix A. M >= 0.
*
* N (input) INTEGER
* The number of columns of the matrix A. N >= 0
*
* OFFSET (input) INTEGER
* The number of rows of A that have been factorized in
* previous steps.
*
* NB (input) INTEGER
* The number of columns to factorize.
*
* KB (output) INTEGER
* The number of columns actually factorized.
*
* A (input/output) REAL array, dimension (LDA,N)
* On entry, the M-by-N matrix A.
* On exit, block A(OFFSET+1:M,1:KB) is the triangular
* factor obtained and block A(1:OFFSET,1:N) has been
* accordingly pivoted, but no factorized.
* The rest of the matrix, block A(OFFSET+1:M,KB+1:N) has
* been updated.
*
* LDA (input) INTEGER
* The leading dimension of the array A. LDA >= max(1,M).
*
* JPVT (input/output) INTEGER array, dimension (N)
* JPVT(I) = K <==> Column K of the full matrix A has been
* permuted into position I in AP.
*
* TAU (output) REAL array, dimension (KB)
* The scalar factors of the elementary reflectors.
*
* VN1 (input/output) REAL array, dimension (N)
* The vector with the partial column norms.
*
* VN2 (input/output) REAL array, dimension (N)
* The vector with the exact column norms.
*
* AUXV (input/output) REAL array, dimension (NB)
* Auxiliar vector.
*
* F (input/output) REAL array, dimension (LDF,NB)
* Matrix F**T = L*Y**T*A.
*
* LDF (input) INTEGER
* The leading dimension of the array F. LDF >= max(1,N).
*
* Further Details
* ===============
*
* Based on contributions by
* G. Quintana-Orti, Depto. de Informatica, Universidad Jaime I, Spain
* X. Sun, Computer Science Dept., Duke University, USA
*
* Partial column norm updating strategy modified by
* Z. Drmac and Z. Bujanovic, Dept. of Mathematics,
* University of Zagreb, Croatia.
* -- April 2011 --
* For more details see LAPACK Working Note 176.
* =====================================================================
*
* .. Parameters ..
REAL ZERO, ONE
PARAMETER ( ZERO = 0.0E+0, ONE = 1.0E+0 )
* ..
* .. Local Scalars ..
INTEGER ITEMP, J, K, LASTRK, LSTICC, PVT, RK
REAL AKK, TEMP, TEMP2, TOL3Z
* ..
* .. External Subroutines ..
EXTERNAL SGEMM, SGEMV, SLARFG, SSWAP
* ..
* .. Intrinsic Functions ..
INTRINSIC ABS, MAX, MIN, NINT, REAL, SQRT
* ..
* .. External Functions ..
INTEGER ISAMAX
REAL SLAMCH, SNRM2
EXTERNAL ISAMAX, SLAMCH, SNRM2
* ..
* .. Executable Statements ..
*
LASTRK = MIN( M, N+OFFSET )
LSTICC = 0
K = 0
TOL3Z = SQRT(SLAMCH('Epsilon'))
*
* Beginning of while loop.
*
10 CONTINUE
IF( ( K.LT.NB ) .AND. ( LSTICC.EQ.0 ) ) THEN
K = K + 1
RK = OFFSET + K
*
* Determine ith pivot column and swap if necessary
*
PVT = ( K-1 ) + ISAMAX( N-K+1, VN1( K ), 1 )
IF( PVT.NE.K ) THEN
CALL SSWAP( M, A( 1, PVT ), 1, A( 1, K ), 1 )
CALL SSWAP( K-1, F( PVT, 1 ), LDF, F( K, 1 ), LDF )
ITEMP = JPVT( PVT )
JPVT( PVT ) = JPVT( K )
JPVT( K ) = ITEMP
VN1( PVT ) = VN1( K )
VN2( PVT ) = VN2( K )
END IF
*
* Apply previous Householder reflectors to column K:
* A(RK:M,K) := A(RK:M,K) - A(RK:M,1:K-1)*F(K,1:K-1)**T.
*
IF( K.GT.1 ) THEN
CALL SGEMV( 'No transpose', M-RK+1, K-1, -ONE, A( RK, 1 ),
$ LDA, F( K, 1 ), LDF, ONE, A( RK, K ), 1 )
END IF
*
* Generate elementary reflector H(k).
*
IF( RK.LT.M ) THEN
CALL SLARFG( M-RK+1, A( RK, K ), A( RK+1, K ), 1, TAU( K ) )
ELSE
CALL SLARFG( 1, A( RK, K ), A( RK, K ), 1, TAU( K ) )
END IF
*
AKK = A( RK, K )
A( RK, K ) = ONE
*
* Compute Kth column of F:
*
* Compute F(K+1:N,K) := tau(K)*A(RK:M,K+1:N)**T*A(RK:M,K).
*
IF( K.LT.N ) THEN
CALL SGEMV( 'Transpose', M-RK+1, N-K, TAU( K ),
$ A( RK, K+1 ), LDA, A( RK, K ), 1, ZERO,
$ F( K+1, K ), 1 )
END IF
*
* Padding F(1:K,K) with zeros.
*
DO 20 J = 1, K
F( J, K ) = ZERO
20 CONTINUE
*
* Incremental updating of F:
* F(1:N,K) := F(1:N,K) - tau(K)*F(1:N,1:K-1)*A(RK:M,1:K-1)**T
* *A(RK:M,K).
*
IF( K.GT.1 ) THEN
CALL SGEMV( 'Transpose', M-RK+1, K-1, -TAU( K ), A( RK, 1 ),
$ LDA, A( RK, K ), 1, ZERO, AUXV( 1 ), 1 )
*
CALL SGEMV( 'No transpose', N, K-1, ONE, F( 1, 1 ), LDF,
$ AUXV( 1 ), 1, ONE, F( 1, K ), 1 )
END IF
*
* Update the current row of A:
* A(RK,K+1:N) := A(RK,K+1:N) - A(RK,1:K)*F(K+1:N,1:K)**T.
*
IF( K.LT.N ) THEN
CALL SGEMV( 'No transpose', N-K, K, -ONE, F( K+1, 1 ), LDF,
$ A( RK, 1 ), LDA, ONE, A( RK, K+1 ), LDA )
END IF
*
* Update partial column norms.
*
IF( RK.LT.LASTRK ) THEN
DO 30 J = K + 1, N
IF( VN1( J ).NE.ZERO ) THEN
*
* NOTE: The following 4 lines follow from the analysis in
* Lapack Working Note 176.
*
TEMP = ABS( A( RK, J ) ) / VN1( J )
TEMP = MAX( ZERO, ( ONE+TEMP )*( ONE-TEMP ) )
TEMP2 = TEMP*( VN1( J ) / VN2( J ) )**2
IF( TEMP2 .LE. TOL3Z ) THEN
VN2( J ) = REAL( LSTICC )
LSTICC = J
ELSE
VN1( J ) = VN1( J )*SQRT( TEMP )
END IF
END IF
30 CONTINUE
END IF
*
A( RK, K ) = AKK
*
* End of while loop.
*
GO TO 10
END IF
KB = K
RK = OFFSET + KB
*
* Apply the block reflector to the rest of the matrix:
* A(OFFSET+KB+1:M,KB+1:N) := A(OFFSET+KB+1:M,KB+1:N) -
* A(OFFSET+KB+1:M,1:KB)*F(KB+1:N,1:KB)**T.
*
IF( KB.LT.MIN( N, M-OFFSET ) ) THEN
CALL SGEMM( 'No transpose', 'Transpose', M-RK, N-KB, KB, -ONE,
$ A( RK+1, 1 ), LDA, F( KB+1, 1 ), LDF, ONE,
$ A( RK+1, KB+1 ), LDA )
END IF
*
* Recomputation of difficult columns.
*
40 CONTINUE
IF( LSTICC.GT.0 ) THEN
ITEMP = NINT( VN2( LSTICC ) )
VN1( LSTICC ) = SNRM2( M-RK, A( RK+1, LSTICC ), 1 )
*
* NOTE: The computation of VN1( LSTICC ) relies on the fact that
* SNRM2 does not fail on vectors with norm below the value of
* SQRT(DLAMCH('S'))
*
VN2( LSTICC ) = VN1( LSTICC )
LSTICC = ITEMP
GO TO 40
END IF
*
RETURN
*
* End of SLAQPS
*
END
|
